Meet Kai
Intelligence you can defend.
Kai drafts the assessment, surfaces the lesson you'd have missed, and flags the gap before it ships. It prepares the record in your voice — and stops short of the decision. You decide; Kai never does.
Assessment — segregation of duties
KAI IS DRAFTING…▋
Three things Kai does. One it doesn't.
Kai is defined by its conduct, not a personality. It drafts, surfaces and flags. The fourth box is reserved for a human — and it always will be.
Drafts
"Control appears configured but untested. Suggested: schedule a walkthrough before sign-off."
Surfaces
Kai surfaced CHG-0934 — a similar control gap closed 11 months ago, with the remediation that worked.
Flags
"Flag: required control appears missing from the approver chain — confirm before this gate is marked effective."
The accept. The sign-off. The closing of the gate. This box has no Kai signature — and never will.
Kai prepares the record. You sign it.
You can always tell who wrote it.
One visual grammar, three states. Over-honest about AI involvement, by design — so a reader never has to ask what Kai touched.
KAI DRAFT
unfinished · dashed hairline · faint petrol tint · no one owns it yet
ACCEPTED · HUMAN-OWNED
resolved · solid border · petrol underline · signed by you
FLAGGED · NEEDS A HUMAN
halted · petrol marker · status chip · blocks on you
THE RECORD IT KEEPS
The thread writes itself.
Every Kai action and every human acceptance lands as one line — actor, act, timestamp. Nothing Kai does is silent, and nothing is anonymous.
It cites. It never asserts.
Kai shows how sure it is and where the obligation comes from. Confidence is a prompt to look closer — not a verdict, and never a green light.
The approver list for CHG-1182 omits the segregation-of-duties reviewer named in your control catalogue.
moderate — a prompt to look, not a conclusion
Kai stops at the decision.
Kai's reasoning is visible end to end — every obligation read, every prior change matched, every check run. The thread is a chain of closed nodes. The last node stays open, because it's yours.
intelligence, made legible —
and terminated at the accountable human.
KAI · REASONING TRACE · CHG-1182
Read DORA Art. 6 — ICT risk-management framework
obligation · source
Matched CHG-0934 — a prior segregation-of-duties gap
precedent · 11 months ago
Checked approver list against your control catalogue
evidence · mismatch found
Drafted the assessment in your house voice
awaiting acceptance
Decision — reserved for you
Kai stops here. The node stays open until a human closes it.
See Kai draft against a real change — and watch it stay in its lane.
We'll run it live on something you're governing.
Request a demo